Anthropic explores in-house AI chip development for Claude models

As demand for advanced AI models continues to grow, Anthropic appears to be expanding beyond model development and into chip design. The US-based AI company has started hiring engineers for a Reinforcement Learning team that could work on custom AI chips designed to improve the performance of its Claude models.

Anthropic recently relaunched its most capable AI model, Claude Mythos, with cybersecurity capabilities, along with Claude Fable 5. The company is now hiring a Research Engineer for its Chip Design RL team at its San Francisco and New York City offices.

According to the job listing, the team works on reinforcement learning research and the development of Claude AI models, with “significant impacts on the autonomy and coding capabilities of Claude Fable 5 and Opus 4.8”.

The new Research Engineer will help “advance” Claude’s ability to design silicon chips. The role also involves developing reinforcement learning environments and evaluations for agentic RTL generation, design verification and physical design optimisation.

The job listing indicates that Anthropic could be preparing to design its own AI chips for Claude models. A company spokesperson has also reportedly confirmed that Anthropic plans to establish an in-house engineering team focused on chip design.

The company is expected to co-design hardware and AI models, which could allow Claude models to “run faster and more efficiently” at the scale required by Anthropic’s customers.

While Anthropic is reportedly planning to handle chip design internally, manufacturing would need to be outsourced. The company is reportedly in discussions with Samsung to produce the custom chips. Samsung’s System LSI Division, which handles chip manufacturing, could potentially build Anthropic’s AI chips in the future.

However, neither Anthropic nor Samsung has officially confirmed the reported manufacturing partnership.